Learning to segment roads for traffic analysis in urban images

Vehicle global 6-DoF pose estimation under traffic surveillance camera

Map matching of sparse vehicle trajectories is a fundamental problem in location-based services, such as traffic flow analysis and vehicle routing. Existing literature mainly relies on sequence-to-sequence (Seq2Seq) models to capture the intra-trajectory correlation of an input trajectory and to sequentially predict the matched road segments. Due to the limited expressive capability of sequential models, these methods fall short of extracting inter-trajectory and trajectory-road correlations as well as correlation between road segments. We present <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">GraphMM</b> , a graph-based approach that explicitly utilizes all aforementioned correlations. Our model exploits the graph nature of map matching and incorporates graph neural networks and conditional models to leverage both road and trajectory graph topology, while manages to align road segments and trajectories in latent space. We formally analyze the expressive power of our model in capturing various correlations and propose efficient algorithms for model training and inference. In particular, our optimization techniques dramatically reduce the computational complexity, making our model feasible on datasets with thousands of road segments. Extensive experiments show that our model significantly enhances prediction accuracy, while improving training and inference efficiency by up to an order of magnitude over both the industrial implementation of the hidden Markov model and state-of-the-art Seq2Seq-based methods.

In urban mobility, map-matching aims to project GPS points generated by moving objects onto the road segments representing the actual object positions. Up to now, map-matching has found interesting applications in traffic analysis, frequent path extraction, and location prediction. However, state-of-art implementations of map-matching algorithms are either private, sequential or inefficient. In this paper, we propose an extension of an existing serial algorithm of known efficiency by reformulating it in a distributed way, in order to achieve great scalability on real big data scenarios. Furthermore, we enhance the robustness of the algorithm, which is based on a first order Hidden Markov Model, by introducing a smart strategy to avoid gaps in the matched road segments; indeed, this problem may occur under sparse GPS sampling or in urban areas with highly fragmented road segments. Our implementation is based on Apache Spark and is publicly available on Github. The implementation is tested against a dataset with 7.8 million GPS points in Milan.

Road detection is a crucial problem for autonomous navigation system (ANS) and advance driver-assistance system (ADAS). In this paper, we propose a hierarchical road detection method for robust road detection in challenging scenarios. Given an on-board road image, we first train a Gaussian mixture model (GMM) to obtain road probability density map (RPDM), and next oversegment the image into superpixels. Based on RPDM and superpixels, initial seeds are selected in an unsupervised way, and the seed superpixels iteratively try to occupy their neighbors according to GrowCut framework, the road segment is obtained after convergency. Finally, we refine the road segment with a conditional random field (CRF), which enforces the shape prior on the road segmentation task. Experiments on two challenging databases demonstrate that the proposed method exhibits high robustness compared with the state-of-the-art.

Multi-object urban dataset: A resource for detecting pedestrians, traffic and motorbikes

Advances and applications of computer vision techniques in vehicle trajectory generation and surrogate traffic safety indicators

Abstract. Floating car technology is widely applied in traffic information collection and has become a common method for real-time and dynamic monitoring of urban traffic conditions. Using floating car data from Shanghai as the research object, this paper analyzes the positioning error of floating cars and finds that most errors are within 15 meters. The error is significantly affected by vehicle speed but less by time, and the error decreases as the road grade increases. Speed and travel time ratio are used as parameters to assess the spatiotemporal traffic state, exploring the spatial distribution and temporal evolution of speed, as well as the spatiotemporal distribution of traffic congestion. In the spatial distribution of speed, clustering analysis is used to deduce the temporal distribution pattern of speed on the same road segment. Based on this, corresponding traffic management measures can be applied to different road segments and at different times.

Road detection is an essential component of field robot navigation systems. Vision sensors play an important role in road detection for their great potential in environmental perception. In this paper, we propose a hierarchical vision sensor-based method for robust road detection in challenging road scenes. More specifically, for a given road image captured by an on-board vision sensor, we introduce a multiple population genetic algorithm (MPGA)-based approach for efficient road vanishing point detection. Superpixel-level seeds are then selected in an unsupervised way using a clustering strategy. Then, according to the GrowCut framework, the seeds proliferate and iteratively try to occupy their neighbors. After convergence, the initial road segment is obtained. Finally, in order to achieve a globally-consistent road segment, the initial road segment is refined using the conditional random field (CRF) framework, which integrates high-level information into road detection. We perform several experiments to evaluate the common performance, scale sensitivity and noise sensitivity of the proposed method. The experimental results demonstrate that the proposed method exhibits high robustness compared to the state of the art.

Batch classifier with adaptive update for backbone traffic classification

For the performance evaluation of future wireless communication systems, such as the fourth generation wireless networks, traffic traces of realistic services are needed. Multiple description coding (MDC) is gaining a lot of interest lately and is a viable solution to provide robust video services over single or multi hop wireless networks and MDC introduces more flexibility for network coding. Furthermore it has the capability to support heterogeneous terminals as they are accepted to be used in 4G wireless networks. By means of twelve well known video sequences, in different video formats, we generate the frame size traces and evaluate the multiple description coding characteristics. In addition to that we highlight the expected overhead due to the underlying RTP/UDP/IP protocol suite. As an objective quality measurement at the application layer, we investigate the video quality in dependency of lost and error-prone descriptors. This allows researcher to convert the network losses of their network models directly into video quality values. This step makes the work unique as single layer coded video streams would always need further postprocessing to retrieve the video quality.

In this paper we propose a traffic surveillance camera calibration method based on detection of pairs of vanishing points associated with vehicles in the traffic surveillance footage. To detect the vanishing points we propose a CNN which outputs heatmaps in which the positions of vanishing points are represented using the diamond space parametrization which enables us to detect vanishing points from the whole infinite projective space. From the detected pairs of vanishing points for multiple vehicles in a scene we establish the scene geometry by estimating the focal length of the camera and the orientation of the road plane. We show that our method achieves competitive results on the BrnoCarPark dataset while having fewer requirements than the current state of the art approach.

In this work, a vehicle tracking system is developed to deal with daytime and nighttime traffic surveillance videos. For daytime videos, vehicles are detected via background modeling. For nighttime videos, headlights of vehicles need to be located and paired to initialize vehicles for the tracking purpose. An algorithm based on likelihood computation is developed to pair the headlights of vehicles. In addition, we apply a specialized system state transition model of the Kalman filter to adapt to common settings of traffic surveillance cameras. The experimental results have shown that the proposed method can effectively track vehicles in both daytime and nighttime surveillance videos.

This paper studied air traffic complexity by the theory of complex analysis. The calculation system of air traffic complexity measure and the robustness analysis of air traffic flow were proposed. When the aircraft influenced the air traffic flow complexity mostly, the flight warning system began to work. This paper discussed the theory of the monitor- ing and early warning based on the complexity of air traffic complexity measure analysis. The studies shown that the method provided in this paper could discern the state parameter associated through technical indicators from complex, un- certain data in the calculation of system complexity, and the research also pointed out the flights which had larger effects on the air traffic through robustness change. The researches show that the complexity theory proposed in this paper can carry on monitoring and early warning on flights through the complexity measure and the air traffic complexity analysis. Aerospace system is a classic complex giant system, which is composed of aerospace structure, air traffic flow consisting of aircraft in operation, air traffic control rules and programs, as well as management and controller' ability. The complexity of air traffic management is up to three factors: 1.The mutual influence among people, equipment, environ- ment and management; 2. Partial but not complete certainty of the system's condition during air traffic operation; 3. The difficulty of measuring data to evaluate the system objective- ly, which needs more than qualitative analysis based on ex- perts' research and scoring as well as flight statistics. The complexity of air traffic is a quantitative result under the interaction between aircraft-consisted air traffic flow and aerospace structure. The management of air traffic complexi- ty is a reliable method to handle comparative intense air traf- fic in order to identify the complicated air traffic condition and take effective actions including operation programs to scale down the complexity to an acceptable range or level with key factors which matter to the detected complexity. Based on that, it is a significant problem that how to quantify the predicted air traffic flow complexity.

In this paper, we describe a system that is capable of detecting and segmenting objects from video frames which helps in traffic surveillance. Shadow is one of the problems faced by most of the object detection systems. It will affect the result of object detection and segmentation. Hence a shadow removal method is applied in the preprocessing phase of the system to amplify the accuracy of detection. MATLAB is the major platform for developing this system. By differentiating the background and foreground of the video scenes using MATLAB embedded functions, the foreground regions are segmented as the objects. The segmented objects are saved for further development of computer vision applications such as object recognition and classification. The proposed system is tested with four traffic video scenes and the experimental results show that the system works well with an accuracy of approximately 90% achieved.

Background subtraction in real applications: Challenges, current models and future directions